nUGV-1UAV robot swarms: low-altitude remote sensing-based decentralized planning framework in-field environments

Abstract

Hybrid-rice seed production demands rapid removal of heterologous plants. We present a decentralized nUGV-1UAV framework that couples low-altitude remote sensing with on-board swarm planning to accomplish this task in large paddy fields. A single UAV performs one-off high-resolution mapping; thereafter, multiple UGVs rely solely on the downloaded map and peer-to-peer communication to execute impurity removal. A topology-guided hybrid A* planner generates homotopy-consistent routes, while a decoupled space–time optimizer refines trajectories for curvature and collision constraints. Field experiments covering 12.7 acres with 73 impurity targets show that a fleet of six UGVs finishes the task in 1.21 h, attaining an individual UGV efficiency of 6 989 m²/h (≈10.5 acres/h). The optimal UGV-to-impurity ratio is 0.47: 5.75: 1 (UGV: impurities: acre). Simulations up to 200 acres demonstrate linear scalability with <5 % deviation from the analytical model. Even when the UAV is disabled, UGVs maintain 92 % task completion using offline maps, confirming robust decentralization.